On 09/08/18 03:41, Danil Iashchenko wrote:
> Add erosion_opencl, dilation_opencl filters. Behave like existing erosion and
> dilation filters.
> ---
>
> Thanks! Fixed.
>
> configure | 2 +
> libavfilter/Makefile | 4 +
> libavfilter/allfilters.c | 2 +
> libavfilter/opencl/neighbor.cl | 87 +++++++++++
> libavfilter/opencl_source.h | 1 +
> libavfilter/vf_neighbor_opencl.c | 327
> +++++++++++++++++++++++++++++++++++++++
> 6 files changed, 423 insertions(+)
> create mode 100644 libavfilter/opencl/neighbor.cl
> create mode 100644 libavfilter/vf_neighbor_opencl.c
>
> ...
> diff --git a/libavfilter/vf_neighbor_opencl.c
> b/libavfilter/vf_neighbor_opencl.c
> new file mode 100644
> index 0000000..7ba759b
> --- /dev/null
> +++ b/libavfilter/vf_neighbor_opencl.c
> ...
> +static int neighbor_opencl_make_filter_params(AVFilterContext *avctx)
> +{
> + NeighborOpenCLContext *ctx = avctx->priv;
> + int *matrix = NULL;
> + size_t matrix_bytes;
> + cl_mem buffer;
> + cl_int cle;
> + int i;
> +
> + for (i = 0; i < 4; i++) {
> + ctx->threshold[i] /= 255.0;
> + }
> + for (i = 0; i < 9; i++) {
> + matrix_bytes = sizeof(float)*9;
sizeof(int)?
> + matrix = av_malloc(matrix_bytes);
> + }
I don't think you want a loop around this? It allocates extra instances of
matrix which never get freed.
Since it's small it might be easier to just put "cl_int matrix[9]" on the stack.
> + if (!matrix) {
> + av_freep(&matrix);
> + return AVERROR(ENOMEM);
> + }
> + matrix[4] = 0;
> + for (i = 0; i < 8; i++) {
> + if (ctx->coordinates & (1 << i)) {
> + matrix[i > 3 ? i + 1: i] = 1;
> + }
> + }
> + buffer = clCreateBuffer(ctx->ocf.hwctx->context,
> + CL_MEM_READ_ONLY |
> + CL_MEM_COPY_HOST_PTR |
> + CL_MEM_HOST_NO_ACCESS,
> + matrix_bytes, matrix, &cle);
This buffer is never released.
> + if (!buffer) {
> + av_log(avctx, AV_LOG_ERROR, "Failed to create matrix buffer: "
> + "%d.\n", cle);
> + av_freep(&matrix);
> + return AVERROR(EIO);
> + }
> + ctx->coord = buffer;
> + av_freep(&matrix);
> +
> + return 0;
> +}
> ...
Tested and has identical output to existing dilation/erosion filters, so
otherwise LGTM.
Thanks,
- Mark
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
